Browsing by Author "Louvieris, Panos"

Now showing 1 - 2 of 2
  • Thumbnail Image
    ItemOpen Access
    Eliciting expert knowledge to inform training design
    (Association for Computing Machinery (ACM), 2019-09-10) Clewley, Natalie; Dodd, Lorraine; Smy, Victoria; Witheridge, Annamaria; Louvieris, Panos
    To determine the elicitation methodologies best placed to uncover and capture the expert operator’s reflective cognitive judgements in complex and dynamic military operating environments (e.g., explosive ordinance disposal) in order to develop the specification for a reflective eXplainable Artificial Intelligence (XAI) agent to support the training of domain novices. Approach: A bounded literature review of the latest developments in expert knowledge elicitation was undertaken to determine the ’art-of-the-possible’ in respects to uncovering an expert’s cognitive judgements in complex and dynamic environments. Candidate methodologies were systematically and critically reviewed in order to identify the most promising methodologies for uncovering expert situational awareness and metacognitive evaluations in pursuit of actionable threat mitigation strategies in high-risk contexts. Research outputs are synthesized into an interview protocol for eliciting and understanding the in-situ actions and decisions of experts in high-risk, complex operating environments. Practical implications: Trainees entering high-risk operating environments can benefit from exposure to expert reflective strategies whilst learning the trade. Typical operator training focuses on technical aspects of threat mitigation but often overlooks reflective self-evaluation. The present study represents an initial step towards determining the feasibility of designing a reflective XAI agent to augment the performance of trainees entering high-risk operations. Outputs of the expert knowledge elicitation protocol documented here shall be used to refine a theoretical framework of expert operator judgement, in order to determine decision support strategies of benefit to domain novices.
  • Thumbnail Image
    ItemOpen Access
    A Markov multi-phase transferable belief model for cyber situational awareness
    (IEEE, 2019-02-06) Ioannou, Georgios; Louvieris, Panos; Clewley, Natalie
    eXfiltration Advanced Persistent Threats (XAPTs) increasingly account for incidents concerned with critical information exfiltration from High Valued Targets (HVTs). Existing Cyber Defence frameworks and data fusion models cannot cope with XAPTs due to a lack of provision for multi-phase attacks characterized by uncertainty and conflicting information. The Markov Multi-phase Transferable Belief Model (MM-TBM) extends the Transferable Belief Model to address the multi-phase nature of cyber-attacks and to obtain previously indeterminable Cyber SA. As a data fusion technique, MM-TBM constitutes a novel approach for performing hypothesis assessment and evidence combination across phases, by means of a new combination rule, called the Multi-phase Combination Rule with conflict Reset (MCR 2 ). The impact of MM-TBM as a Cyber Situational Awareness capability and its implications as a multi-phase data fusion theory have been empirically validated through a series of scenario-based Cyber SA experiments for detecting, tracking, and predicting XAPTs.